Why I make such a big deal about gain structure

[agedhorse]

Because it's important, and in fact it's critical to understand the gain structure of each piece of equipment and the equipment it will be interfaced with. The system is only as good as it's weakest link and I think there is such a lack of understanding about this topic that it deserves walking through an example.

 

Some of you know that I have designed products for this industry for ~25 years. Because of this I am intimately familar with the technical capabilities of a lot of different pieces of gear, sometimes to the point of a little too much. I also get a few comments here about how I don't understand how the real world applications work... specifically acoustic analysis and more recentlt gain structure and DSP/speaker processor program development. Well, hogwash. I do this stuff every day full time.

 

I thought I would share with you guys WHY understanding the minute details of gain structure is important, and what can happen when you think you know more than you do.

 

I received a sample of a DSP that I thought might have some value as an entry level product for some applications where industry standard adherance was not necessary, especially something in the 4 in/8 out catagory. The DSP was a BBE DS-48, made by an OEM company in China (I forget the name of the company, but Mark and I saw them a few years ago and recognized the unit before BBE branded them as their own here in the US.

 

I started measuring things and did not get numbers that I expected, not did I get numbers that were particularly useable. The input overloaded and distorted at about +7 or 8dBu, and the output was unable to attain +8dBu before clipping or limiting regardless of the programming of the input module's gain setting. The meters indicated clipping before clipping actually occured and the limit thereshold of +15dB was clearly at odds with the measured numbers.

 

Why is it important to understand these numbers?

 

1. The maximum output of a typical pro audio console is about +22 to +26dBu before clipping. This means that it's entirely possible to drive the DSP well into clipping but have a false indication of what is really happening because the meter is not accurate.

 

2. With a maximum output level of the DSP being only +7 dBu, there are power amps that this DSP unit can not really drive to rated power before clipping. There is also a descrepency between the limiting module and the actual limiting, meaning that you can limit at +8dBu depending on the level of the input signal, but that may not be appropriate. Some amps that have a >+7dBu input sensitivity, amps such as the QSC PL-3's (on 26dB setting and the 380 on the 32dB setting), many of the Crest Pro series products, the larger Peavey IPR's, etc.) Thisnis a real issue, as almost all amps on the market will need to be operated at 100% sensitivity.

 

3. With a lack of input and output headroom, there is insufficient overall headroom to allow for balancing outputs with the I/O gain modules and still be able to drive the amps to rated output. Say for the common example where subs need to be say 4dB hotter than the tops, there is not 4dB of surplus dynamic range to work with in order to be able to achieve this without overloading or unintentionally limiting.

 

I do not see how folks using this product do not see these problems. They are so obvious that within 5 minutes I realized there were some severe shortcomings in this product. In fact, it's pretty unusable for any even remotely critical application.

 

Now, I explored the internals ("guts" for you manly men) and looked at what the limiting factors were. Of course there are no schematics available, but the I/O topology is quite familiar to those of us who design this suff, so a little hunting down and it turns out that the problem on the input side was their designer's choice of A/D scaling. That's an easy modification and one resistor sets the scaling so I will decrease the input scale by 8dB, meaning that now the input overload is +7dBu plus 8dB = +15dBu (a little over 6 volts RMS) which happens to be the same as the input overload of the GX series amps, and also the minimum that I feel is reasonable or justifiable in pro audio.

 

The output is another problem entirely, and fortunately the processor's analog audio path is supported by a +/-15V power supply, so +22dBu maximum output level in bridge mode is attainable but what to do... Looking at the output of the D/A, it is limited to about 4.5 volts peak-peak (typical) but there is very limited gain in the make-up amp and this too must be increased by at least 8dB for a maximum differential drive level of +15dB based on the limitations placed by the D/A.

 

Ok, the metering next. What happens when we mess with the scaling. Turns out, all of the metering has nothing to do with the real signal levels but has to do with the internal DSP levels, and the meter scale is relative to the approximate full sacle of the DSP. This means that it will read and track accurately all of the scaling, so that DSP (or A/D) clip now reads clip on the input meter, which translates to +15dBu, and output clip on the DSP (or D/A) also translates to full scale on the D/A and happens to be +15dBu or 6 volts RMS.

 

Now, this processor is fully compatable with all of the pro audio consoles (you can still clip the input of the ADC, but it's at +15dBu which is acceptable knowing this limit) and can drive the inputs of all power amps to rated power and still have headroom to allow for balancing the different bands with different signal levels and not run out of drive.

 

This is very typical of DSPs in many power amps, and some inexpensive consoles. It pays to really know exactly what you are doing AND VERIFYING before assuming you understand how something works. Learn the math, or at least learn the concepts so you have an idea of what to look for when something is not cutting it. It also may help some of you folks who doubt my comments and recommendations, arguing is not going to change how something really works.

 

I don't know yet how these modifications will affect the noise floor, and how they will affect the limiter calibration. I hope the noise floor changes will not be an issue, and I also hope that my choice of values will make the limiter's _15dB scale match up pretty closely with reality (think there might have been a reason why 15dBu was my chosen target?

 

I also do not know if it will be worth the cost of modifying, the problem is that it also (I assume ) will void the warranty. That doesn't bother me but really could be an issue in the field for a typical user.

 

Hope this causes some serious thought for at least some of you.

[Dogoth]

Is this a new product? About 2 years ago I was shopping for a multi channel crossover (really just to simply biamp multiple channels of 2 way monitors). I came across a BBE product (I don't remember the model number) who's description was suspiciously close. It SEEMED cost effective and I almost bought it but I really didn't need all of the bells & whistles and (based upon information from this forum) didn't really trust the brand name. I settled on an Ashley 4 in 8 out analog crossover. It has been in continuous use since and will probably last for years. It sounds like I made the right choice.

 

Nice review. Thanks. Unfortunatly we don't all have the test equipment to find these problems out ahead of time. We usualy wait until that one day you have to push your system to it's limits and then find out the hard way (that's usualy about 4 days after the return policy is up :-).

[Shaster]

Wow! That's gonna take a while, no a long while to fully digest, but I hope I've got the gist of what you're saying.

 

It points to something I was posting about a while back.... when we see clip lights, how have they been calibrated, how accurate, and to what measurement? I sometimes feel clip lights are either well before clip, or sometimes right at it - depending on the product and how it's been set up.

 

Fascinating stuff, and your topic can certainly explain certain anomalies that crop up from time to time.

 

Thanks for taking the time to post this.

[tuchan]

Nice read, very informative, thankyou.

[oldschooler]

Very interesting because that unit got some praise from the guys at Prosoundweb as being usable. I believe it comes with a 2 or 5 yr warranty also making it seem like a decent product. I will admit I looked at it at one point but decided to stick with the Driverack 260.

[Dookietwo]

"""""""Ok, the metering next. What happens when we mess with the scaling. Turns out, all of the metering has nothing to do with the real signal levels but has to do with the internal DSP levels, and the meter scale is relative to the approximate full sacle of the DSP. This means that it will read and track accurately all of the scaling, so that DSP (or A/D) clip now reads clip on the input meter, which translates to +15dBu, and output clip on the DSP (or D/A) also translates to full scale on the D/A and happens to be +15dBu or 6 volts RMS.""""""

 

The output metering is dependant on the limiter settings in DBV. Note the "V".

If you set the limiter to 3.5 DBV the meters will reach full scale when 3.5 DBV is at the outputs.

Set the limiter to -1 DBV and the meters will reach full scale at -1 DBV. The metering and clip, limiter on light are dependant on this setting.

This lets you use full scale metering regardless of what amp your using.

This is covered in the manual. If you don't set the limiter to the input sen. in DB"V" of the amp you want to drive then you won't get correct metering.

 

The BBE crossover is not a greatest unit in the world but its not bad. I have one for my home stereo that works fine. From time to time I've pulled it out for live sound use. Knowing as an informed user of its metering and max input/output range I find it easy to run amps that require 3.5 to 4 DB"V" to reach full power.

 

There is 2 basic points of view as far as gain staging.

1 - is to run the board up to its max output. 22db "u" as you stated. Then match/ turn down gain for the rest of the system for it to clip/reach full output as well.

2- is to have you board at around +4dbu and have the rest of the system reach full power gain matching past the board.

 

I like No. 2 for my system.

Again using setup 2 I've had no problems with proper gain structure running the BBEs. I've seen installs with them that work fine. I do run the input when doing sound at +6 dbs but it doesn't seem to increase system noise enough to be a problem.

"EDIT" It seems BBE had +4 dbu in mind when they set up this product. Although a person can turn the input gains down for unity. I do agree with Andy. These units would be a great match for QSC type amps.

 

As far as input/output clip I have a unit here at home. I may go through it again this weekend and do a write up on it. I'm pretty sure with a 0db"u" input it will drive an amp with a 3.5 DBV sen. to full power no problems. ( I've done this many times ) I realize 8 dbs of input headroom is not the greatest but its not unuseable. Like every sound product out there if a user doesn't understand how to correctly use the product then it won't give correct results. Not the fault of the product. Just the fault of the user.

 

 

Take a look at the limiter section in the manual. It may help give you better metering results or explain the reason for early output meter clipping you were seeing.

 

Andy. Could you check unity through the unit?

Does a 0dbu input equal 0 dbu on the outputs? Of course no eq or gain changes. I believe the outputs are a little low but never checked it. One of the reason for setting input gain at +6.

 

Doug

[jlb]

Thanks for the good read Aged!

[RoadRanger]

There is 2 basic points of view as far as gain staging. 1 - is to run the board up to its max output. 22db "u" as you stated. Then match/ turn down gain for the rest of the system for it to clip/reach full output as well.

2- is to have you board at around +4dbu and have the rest of the system reach full power gain matching past the board.

3- Open up the amps 100% and balance the system with the output gains of the crossover - your amp's clip limiters won't work if you can't overdrive the amp without clipping something upstream.

[The Chinese]

I find that the best thign is to get the best pieceof gear that you can afford, and research it well. There's NO way in hell that I can match your level of expertise, but I've done OK, by looking & buying gear that has a track record of success, and learning as much as I can on the fly. I've blown one driver in 5 years, and I attribute that to not having a really good RMS limiting scheme in place. It's not about buying cheap & features all the time; sometimes it's better to do one thing well than do a hunderd things badly.

[agedhorse]

3- Open up the amps 100% and balance the system with the output gains of the crossover - your amp's clip limiters won't work if you can't overdrive the amp without clipping something upstream.

Note that I was referencing everything to +4dBu (my test set is programmed so that 0dB is referenced to (offset) +4dBu for quick measurements of multiple points relative to a basis of "zero", so all of my dB comments are off by 4dB, or 4dB low. Otherwise, everything is still accurate but 4dB better than I stated.

 

Exactly Ron, and there's little room for error as well.

 

Doug - the specifications for the limiter in the manual state limiter threshold range of -40 to +3dB (no distinction of reference), the module description (pg. 12) states -20 to +15dBV which the output can not attain, does not match in the slightest the on-board menu for the limiter module. There is no way +15dB threshold referenced to anything correctly since the unit clips/distorts the output at +11 dBu (7dBu I measured with my reference offset by 4dB) and if dBV is used, the unit clips at 9.5dBV. I understand that the unit is workable with some amps, especially those that limit/clip at .775V (0dBu), but there are some amps on the market that need 3.5-5 volts RMS to reach rated power. These are the ones that are going to have troubles.

 

It's not the matter of a user not understanding, it's the point that the product is (unintentionally) misleading, the limiter threshold (an important spec) does not match between the unit menu and the manual specs, there is some kind of limiting on the input that happens between between the input gain adjust and the A/D, bring the input up to the overload point and adjust the gain up several dB and there is no change in the output and the input meter is several dB into clip indication, stuff like this that makes me reiterate the need to stay away from the edge, which is where the basic operating parameters put you to begin with.

 

All in all, if they reworked a few things (specifically the I/O levels, the module menu references and the manual) this could be a good option. The user interface on-board sucks, but maybe the GUI is better.

[agedhorse]

[quote=Dookietwo;42862206Andy. Could you check unity through the unit?

Does a 0dbu input equal 0 dbu on the outputs? Of course no eq or gain changes. I believe the outputs are a little low but never checked it. One of the reason for setting input gain at +6.

 

Doug

 

Ran a few more tests:

 

Unity reads 4.16dB low on the outputs relative to the input. Output meters also read low.

 

Limiter indicators are also off by maybe 4-5dB, indicating limiting before actual limiting occurs.

 

Discovered another anomoly with the metering, when you enter output limiting steady state, the input meter drops 6dB then increases back to the reference reading every second or two. The audio signal levels do not change. There is something very goofy with the metering.

 

The more I mess with this, the more I see an unfinished or not thoroughly tested design.

 

Seems like a pretty good idea with a mediocre execution. I do not see any of these problems with the DBX DR products that I rork with regularly. And I so wanted to like it too.

 

edit: unmuted outputs 7 and 8 and all LEDs came on but there was no signal present at the output jacks. Rebooted unit, the meters returned to normal operation and no cycling of the input meter. Possible that I disrupted something in measurement process, but you can be sure I will keep my eyes open on this.

[RoadRanger]

I think I'll stick with my 'ringer DSP .

[stangconv]

^^^ me too!

 

Ugh the more I read the more I realize I don't know... Almost depressing

[RoadRanger]

Ugh the more I read the more I realize I don't know... Almost depressing

Yah, we can't expect everyone to get a college education just to run sound . That's why I generally recommend amps matched to the speakers run WFO from a DBX223XL and a Rane MQ302 stereo EQ. That's what my "B" rig is - why stress your life with all that "huge amps limited down by DSP" crap that all the sound lemmings seem to be into these days ?

 

Of course active speakers with built in crossovers are even better - still need the MQ302 though .

[agedhorse]

I think I'll stick with my 'ringer DSP .

 

Don't fool yourself to thinking everything is as you expect on that either

 

At least they got the input and outout levers correct at +22dBu

[dboomer]

3- Open up the amps 100% and balance the system with the output gains of the crossover ....

 

Nope ... that's backwards as far as gain structure goes.

[witesol]

this is an intriguing product:

http://www.rane.com/ranegain.html

[RoadRanger]

Nope ... that's backwards as far as gain structure goes.

I don't call it "gain structure" - I call it "just works" as in "as far away from anything clipping as possible" .

[Unalaska]

I've seen a few of the BBE units in use, maybe a couple. The last one was used with an OmniSystems line array and their 4ch amps. The last guy I saw loved it and ordered another 6 on ebay for under $200ea. Dunno why, if it doesn't work it doesn't work. Certainly wouldn't work for my P7000 amps which require +7dbu to clip.

[twostone]

So how is all this gain structure work with active cabs and digital mixer?

[dboomer]

I don't call it "gain structure" - I call it "just works" as in "as far away from anything clipping as possible" .

 

Correct ... with with more understanding it can "just work" even better.

 

I think that Andy's opening statement for this thread needs a little clarification.

 

Gain structure is not important at all if #1 you don't get distortion/clipping and #2 the noise doesn't bother you. Now that I said that those two points are usually difficult to achieve without having something close to proper gain structure.

[dboomer]

 

So how is all this gain structure work with active cabs and digital mixer?

 

 

The same everywhere. You just need to be aware of what metrics you are using to call out specs. Digital usually used "dBfs" where "0" is the absolute max. This makes things easy as you just dial down the amount of headroom you want to reserve. Analog systems usually call out some "0" position with the assumption that there is "x" amount of headroom above that zero.

[RoadRanger]

Now that I said that those two points are usually difficult to achieve without having something close to proper gain structure.

I've just never had an amp that wasn't quiet WFO so haven't had to worry about it. Active speakers are something different - especially those that you can crank up hot enough to use a mic on and don't have a pad or "mic/line" switch. Those seem to need to be at 1 o'clock +- . With the crossover and EQ gains flat I like to see the clip limiter lights flicker when the board is at +3 or 4.

[RoadRanger]

So how is all this gain structure work with active cabs and digital mixer?

Some digital mixers have the clip point at 0 instead of the typical analog at +22 (+-) - you just have to learn that the new "0" is at -24 .

[dboomer]

 

I've just never had an amp that wasn't quiet WFO so haven't had to worry about it.

 

 

So there you go ... Like I said, if it's not a problem then it's not a problem. But ... it could be even quieter (which isn't an issue to you but if the system were in a quiet church you may need that extra range.

 

 

Active speakers are something different - especially those that you can crank up hot enough to use a mic on and don't have a pad or "mic/line" switch. Those seem to need to be at 1 o'clock +- . With the crossover and EQ gains flat I like to see the clip limiter lights flicker when the board is at +3 or 4.

 

 

Again, like Andy is explaining above. You need to get each piece of gear in it's correct zone gain-wise. Where the input to the speaker is set is meaningless (or at least less meaningful) if the mixer output is not in the right place.

 

And again, again ... it you can still satisfy my points #1 and #2 above, then you are good to go even if it could get better (as you don't require the benefit ... well at least this time:)